系统工程理论与实践
繫統工程理論與實踐
계통공정이론여실천
SYSTEMS ENGINEERING--THEORY & PRACTICE
2010年
4期
667-672
,共6页
供应链协调%进化博弈论%强化学习(RL)%遗传算法(GA)
供應鏈協調%進化博弈論%彊化學習(RL)%遺傳算法(GA)
공응련협조%진화박혁론%강화학습(RL)%유전산법(GA)
supply chain coordination%evolutionary game theory%reinforcement learning (RL)%genetic algorithm (GA)
供应链协调问题多数基于主从博弈结构建模,但如果研究对象是相对复杂的供应链结构,理论求解主从博弈问题就变得困难.因此从求解一对一的供应链协调问题开始,针对主从博弈问题的特点,利用个体学习的进化博弈仿真手段,设计了经销商利用经验分布的预期随机需求的信念更新模式与最优反应的决策模式,为生产商分別设计了基于强化学习的信念更新模式与基于遗传算法搜索策略空间的决策模式,并将两者有机结合,取得了博弈问题的均衡解并且验证该解与理论求解结果一致,为进一步求解复杂问题提供了新的途径.
供應鏈協調問題多數基于主從博弈結構建模,但如果研究對象是相對複雜的供應鏈結構,理論求解主從博弈問題就變得睏難.因此從求解一對一的供應鏈協調問題開始,針對主從博弈問題的特點,利用箇體學習的進化博弈倣真手段,設計瞭經銷商利用經驗分佈的預期隨機需求的信唸更新模式與最優反應的決策模式,為生產商分別設計瞭基于彊化學習的信唸更新模式與基于遺傳算法搜索策略空間的決策模式,併將兩者有機結閤,取得瞭博弈問題的均衡解併且驗證該解與理論求解結果一緻,為進一步求解複雜問題提供瞭新的途徑.
공응련협조문제다수기우주종박혁결구건모,단여과연구대상시상대복잡적공응련결구,이론구해주종박혁문제취변득곤난.인차종구해일대일적공응련협조문제개시,침대주종박혁문제적특점,이용개체학습적진화박혁방진수단,설계료경소상이용경험분포적예기수궤수구적신념경신모식여최우반응적결책모식,위생산상분별설계료기우강화학습적신념경신모식여기우유전산법수색책략공간적결책모식,병장량자유궤결합,취득료박혁문제적균형해병차험증해해여이론구해결과일치,위진일보구해복잡문제제공료신적도경.
Problems of coordinating supply chain are based on Stackelberg game model, but if research object is complex supply chain, it is difficult to find equilibrium of Stackelberg game, so evolutionary game theory was introduced. According to characteristics of leaders and followers in Stackelberg game model, learning mechanism is designed for each player respectively. An algorithm of reinforcement learning combined with genetic searching is proposed for leaders (manufacturers), and a learning model of best-reply is designed for followers (retailers).
